MRAM runs six times faster than DRAM; can it supercharge your systems?

Many technologies that have been hyped over the last decade are actually coming to fruition. Hybrid cars are now sold retail throughout the country, wireless broadband is becoming available everywhere, and cell phones have the capability of capturing and transporting live audio and video.

On the memory front, IBM and Infineon recently released their first 16-Megabit Magnetoresistive Random Access Memory (MRAM) chip. Magnetoresistant data-storage technology has been around since 1974 and is the basis for computer memory (originally magnetic core memory) and the foundation of storage technology for the hard disk. Manufacturers switched to cheaper, volatile SRAM and DRAM solutions in the early'80s. But now the assembly lines are firing up as major players such as Philips, Motorola, Infineon, and IBM are preparing to release MRAM components into the marketplace. To some this may not seem significant, but MRAM could very well transform computing as we know it.

How MRAM Works

MRAM uses magnetic fields created in a series of millions of nanoscopic power rails rather than electric charges to store memory. By passing current through electrodes on either end of the power rail, the magnetic polarity is set to either 1 or 0. Information is then read by passing another current through the rail and measuring resistance, any rail set to the state of '1' yields a higher resistance than that of '0' (see Figure 1).

Because the spin state of the electron is stored magnetically, MRAM does not require a constant charge of electricity to maintain the polarity of each rail. Lab tests have shown that the state is very stable, even when exposed to shock or other magnetic fields. Also, the magnetic write capability does not degrade at nearly the rate of current solid-state flash technology.

By harnessing the quantum property spin direction of electrons as well as the charge, MRAM allows for the electrons to exist in several states at once. While every electron is set to a spin state of 'up' or 'down', the horizontal quantum spin states are superpositions of the spin-up and spin-down states (see Figure 2).

This type of quantum computing allows for amazing amounts of information to be compressed in a very small area. MRAM is the next step to utilizing electron spin states commercially; scientists are now attempting to apply spin-polarization in semi-conductors. This level of quantum data storage carries with it the implication of storing an exponential amount of information, regardless of the size of the storage device. Of course technology has a long way to go before they develop devices to take advantage of quantum mechanics in all their glory, but the introduction of MRAM into the market signals a giant leap in that direction.

[FIGURE 1 OMITTED]

The Best of All Worlds

MRAM has access times as fast or faster than SRAM and as dense as DRAM without any of the power or heat issues. It retains data settings after power is removed making it non-volatile, like a hard drive minus the disk spin up and seek times, and it doesn't suffer the write degradation that flash devices have. Mobile devices will benefit from the reduced power demands allowing usage for over a week on a single charge.

Probably the most anticipated feature of MRAM will be instant-on computing--all prevalent software applications and data will reside in memory without any boot-up sequence, a user will literally push a button and the computer will power up exactly as it was when turned off. MRAM's durability and low power needs combined with a dose of radiation hardening make it perfect for defense and aerospace technologies and is set to be the primary data storage technology for satellites.

MRAM and Mass Storage

MRAM has obvious applications to replace SRAM in a host of end user devices that utilize quick access memory; but in larger terms, in the area of long term housing and storage of data, will MRAM fit into an ILM scheme, and if so, where? As prices begin to fall and greater rates of data compression are achieved, will MRAM become a threat in the mass storage universe? Certainly, the faster access time will be very desirable compared to disk or tape, with compression rates as great as they are and the durability of MRAM devices, but more complex technologies would be required to meet the capacity demands of the mass storage universe.

[FIGURE 2 OMITTED]

In the short term, MRAM is set to replace the now volitile solid state devices that allieviate I/O bottlenecks and supply hotfiles on large networks. MRAM's stability makes it a powerful contender in any DR/BC contingency plan. Obviously, for any of these larger scenarios to come true costs will have to come down considerably. The reasons that MRAM hasn't gone into commercial production sooner are cost and scale of technology. Motorola and Philips joined forces earlier this year in an effort to develop semiconductor processes that feature sizes of a 100nm or less, paving the way for MRAM's commercially viable production. Look for MRAM to start showing up in handheld devices in late 2004/early 2005 and spread into other applications quickly thereafter.